Munger and Pothoff 2000; Grofman 2004; Hinich and Munger 1997). However, we are not
concerned about the point at which legislators or candidates position themselves along an
ideological continuum. We are instead interested in the “search” process that legislators use in
an attempt to figure out where the median voter might be (e.g., Kollman, Miller, and Page 1992;
1998). If legislators do not know where the median is located, they will search by taking various
positions and observing feedback from constituents through elections and other means, making
“classification errors” in the process. The more uncertainty they have about their constituents’
preferences, the more they make what appear to be inconsistent decisions in Congress.
Further, we offer a theoretical account of behavior that has previously been considered
idiosyncratic or exceptional. Inconsistent legislators are sometimes referred to as mavericks in
the congressional literature or by the media. They are generally dismissed as outliers who do not
fit the model of a typical senator. Instead of being outliers, however, we think these legislators’
maverick behavior can be explained theoretically. Instead of assuming these misclassified votes
are predictive errors that may be noise, we posit there are clear theoretical reasons to expect
some legislators to make more errors than others. Consistency in roll-call voting is an outcome
that we think can be accounted for theoretically, and not something explained through ad hoc
declarations that some senators are simply mavericks.
A THEORY OF IDEOLOGICAL SEARCH WITH LEGISLATOR UNCERTAINTY
We posit that ideological inconsistency can best be explained through a senator’s
uncertainty over the location of the median voter of his or her constituency. In the context of a
two-candidate election campaign, uncertainty about the median voter typically yields predictions
where candidates will make their “best guess” about the location of the median and both
candidates will take the same position (as long as they do not have personal policy preferences
3